skshiraj Goto Github PK

100days_of_ml_code

ML models

11-apply-truncated-svd-on-donors-choose-dataset-

13-try-various-cnn-networks-on-mnist-dataset-

14--lstm-on-donors-choose

17-microsoft-malware-detection

23-facebook-friend-recommendation

24-apparel-recommendation

25-cnn-on-cifr

4-apply-naive-bayes-on-donors-choose-dataset

5-logistic-regression-on-donors-choose-dataset

6-implement-sgd-for-linear-regression

7-apply-svm-on-donors-choose-dataset-

9-apply-random-forests-gbdt-on-donors-choose-dataset

aaic

aaic-case-studies-

appliedaicourse

This repo has all the code files which were created as part of the assignments to complete the Applied AI Course. The credit for code source structure and data goes to the Applied AI team.

apply-random-forests-gbdt-on-donors-choose-dataset-

backpropagation

bagging_confidence_interval

Applied the concept of bagging on boston housing price prediction dataset creating 30 bags, and trained 30 regression DTs, calculated their train MSE score, thentheir MSE score and oob MSE score on the population data, then calculated a 95% confidence interval (C.I.) on doing the previous task 35 times and storing the mse and oob mse scores and using that as a sample, then made 10 samples from this sample to get the population mean and std on 95% confidence interval using the central limit theorem.

call-backs

callback

case-study-donors-choose

Perform end to end analysis on Donors Choose Dataset

cifar_cnn

cluster

clustering

Apply K means Agglomerative DBSCAN clustering algorithms on Donors Choose dataset

clustering-on-graphdata

cnn

cnn-mnist

CNN networks on MNIST dataset

cnn-on-cifar

cnn-on-cifar-10-dataset

A test accuracy of 90.04% was acheived by training the densenet model from scratch.There are no dropouts,no fully connected layers used.There are only 0.8 million parameters in total.